skip to main content
OSTI.GOV title logo U.S. Department of Energy
Office of Scientific and Technical Information

Title: The performance of minisupercomputers: Alliant, Convex, and SCS (Scientific Computer Systems)

Conference ·
OSTI ID:6134750

Recently several computers have appeared whose architectures feature many, if not all, of the characteristics of supercomputers, but whose electronic technology uses less expensive and more readily available components. Such machines have been labeled ''minisupercomputers.'' Currently, this class of computers includes those manufactured by Convex, Alliant, and Scientific Computer Systems (SCS). The strategy of these minisupercomputer vendors is to produce a machine that may not provide as much absolute performance as the high-end supercomputers, but exhibits a comparable or better price-performance ratio. While it is not the goal of this paper to comment on the specific price effectiveness of any particular machine, we believe that a proper evaluation of the minisupercomputers should include price data. This paper presents a detailed performance comparison of the Convex C-1, Alliant FX/8, and SCS-40 minisupercomputers. While the Convex C-1 is functionally similar to the CRAY-1 supercomputer, the SCS-40 actually emulates a CRAY X-MP/2, running the same instruction set and providing CRAY X-MP software. The Alliant series of machines are multiprocessors, containing up to eight processors (called computational elements, or CEs) that can run separate processes or cooperate on a single application. The Alliant system also includes a compiler that automatically attempts to partition a code for concurrent execution on several processors. In the first sections of this paper, the single processor performance of the FX/8 is compared with that of the C-1 and the SCS-40, as well as with a Digital Equipment Corporation (DEC) VAX 8600. In a later section, the parallel processing performance of the FX/8 is described. 21 refs.

Research Organization:
Los Alamos National Laboratory (LANL), Los Alamos, NM (United States)
DOE Contract Number:
W-7405-ENG-36
OSTI ID:
6134750
Report Number(s):
LA-UR-87-2806; CONF-8708148-1; ON: DE87014765
Resource Relation:
Conference: Vector and parallel processing in computational science, Liverpool, UK, 25 Aug 1987; Other Information: Portions of this document are illegible in microfiche products
Country of Publication:
United States
Language:
English